Dexterous hand grasping action control method, electronic device, and storage medium

Abstract

The application provides a dexterous hand grasping action control method, electronic equipment and storage medium, and belongs to the technical field of robot dexterous hands. The method comprises the following steps: receiving a grasping task, and acquiring image information of a target object in the grasping task; determining a target diameter corresponding to the target object by analyzing the image information; comparing the target diameter with a plurality of preset grasping diameters of the dexterous hand, and determining a target grasping mode according to a comparison result, wherein the target grasping mode comprises a fingertip grasping mode, a fingertip and second joint cooperative grasping mode and a full hand grasping mode; and controlling the dexterous hand to execute the grasping task in the target grasping mode, wherein in the grasping process, three-dimensional force data collected by three-dimensional force sensors arranged at different positions of the dexterous hand is monitored to determine a real-time grasping state, and a corresponding grasping adjustment strategy is executed according to the real-time grasping state. The application can improve the grasping success rate and operation adaptability of the dexterous hand when facing objects of different sizes.

Description

Technical Field

[0001] This application relates to the field of robot dexterous hand technology, and more particularly to a dexterous hand grasping motion control method, electronic device and storage medium. Background Technology

[0002] With the rapid development of robotics technology, the importance of dexterous hands, as key end effectors for physical interaction between robots and their environment, is becoming increasingly prominent. Beyond traditional fields such as industrial assembly and logistics sorting, dexterous hands are also demonstrating enormous potential in high-precision fields like surgical robots, and are expected to mimic or even surpass the dexterity and perception of human hands to perform delicate grasping tasks. The core of achieving this goal lies in the dexterous hand possessing precise perception and intelligent control capabilities regarding the grasping process.

[0003] Currently, the grasping control strategies of robotic dexterous hands in these scenarios, especially in surgical robots, still mainly rely on preset programs or vision-based single action patterns. For example, after determining the grasping point through image recognition, a grasping action with a fixed trajectory is executed. Although some systems have introduced haptic feedback, it is usually only used for simple force control or anti-slip detection, and it fails to effectively fuse multi-source sensor information, resulting in low adaptability and success rate in grasping objects of different sizes.

[0004] Therefore, there is an urgent need to propose a dexterous hand grasping motion control method that can adapt to task targets of different sizes. Summary of the Invention

[0005] The purpose of this application is to provide a dexterous hand grasping motion control method, electronic device, and storage medium to solve the above-mentioned problems.

[0006] To achieve the above objectives, firstly, this application proposes a method for controlling the grasping action of a dexterous hand, the method comprising: Receive a grabbing task and acquire image information of the target object in the grabbing task; By analyzing the image information, the target diameter corresponding to the target object is determined; The target diameter is compared with multiple preset grasping diameters of a dexterous hand, and the target grasping mode is determined based on the comparison results. The target grasping mode includes a fingertip grasping mode, a fingertip and second joint coordinated grasping mode, and a full-hand grasping mode. The dexterous hand is controlled to perform the grasping task in a target grasping mode. During the grasping process, the real-time grasping state is determined by monitoring the three-dimensional force data collected by three-dimensional force sensors arranged on different parts of the dexterous hand, and the corresponding grasping adjustment strategy is executed according to the real-time grasping state.

[0007] In some embodiments, the plurality of preset grasping diameters includes at least a first grasping diameter, a second grasping diameter, and a third grasping diameter; the step of comparing the target diameter with the plurality of preset grasping diameters of the dexterous hand and determining the target grasping pattern based on the comparison result includes: When the target diameter is less than or equal to the first grasping diameter of the dexterous hand, the dexterous hand is controlled to perform the grasping task in a fingertip grasping mode. When the target diameter is greater than the first grasping diameter of the dexterous hand and less than or equal to the second grasping diameter, the dexterous hand is controlled to perform the grasping task in a grasping mode that coordinates the fingertips and the second joint. When the target diameter is greater than the second grasping diameter of the dexterous hand and less than or equal to the third grasping diameter, the dexterous hand is controlled to perform the grasping task in a full-hand grasping mode.

[0008] In some implementations, the controlled dexterous hand performs the grasping task in a fingertip grasping mode, including: Control the dexterous hand to move to the preset grasping position with the thumb and forefinger spread out; The thumb and index finger are controlled to perform a pinching action, and the pinching action of the contacting fingers is stopped when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object. Control other fingers to perform a retraction action towards the target object, and stop performing the retraction action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object; Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

[0009] In some embodiments, the controlled dexterous hand performs the grasping task in a fingertip and second joint coordinated grasping mode, including: Control the dexterous hand to move to the preset grasping position with the second joints of the thumb and index finger spread to the first preset angle and the fingertips spread to the preset limit angle; The second joints of each finger of the dexterous hand are controlled in sequence to perform tightening actions, and the tightening action of the contacting finger is stopped when the three-dimensional force sensor of the corresponding second joint detects the contact force with the target object. The dexterous hand sequentially controls the fingertips of each finger to perform a clamping action, and stops performing the clamping action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object; Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

[0010] In some implementations, the controlled dexterous hand performs the grasping task in a full-hand grasping mode, including: Control the dexterous hand to move toward the target object with all fingers' third joints spread to a second preset angle, and the second knuckles and fingertips spread to a preset limit angle, until the movement stops when the three-dimensional force sensor arranged in the palm detects the contact force; The third joint of each finger of the dexterous hand is controlled to perform a tightening action in sequence, and the tightening action of the contacting finger is stopped when the three-dimensional force sensor of the corresponding third joint detects the contact force with the target object. The second joints of each finger of the dexterous hand are controlled in sequence to perform tightening actions, and the tightening action of the contacting finger is stopped when the three-dimensional force sensor of the corresponding second joint detects the contact force with the target object. The dexterous hand sequentially controls the fingertips of each finger to perform a clamping action, and stops performing the clamping action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object; Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

[0011] In some implementations, determining the real-time grasping state by monitoring three-dimensional force data collected by three-dimensional force sensors arranged at different parts of the dexterous hand, and executing a corresponding grasping adjustment strategy based on the real-time grasping state, includes: By monitoring the three-dimensional force data collected by three-dimensional force sensors placed at different parts of the dexterous hand, the real-time normal resultant force is determined; When the real-time normal resultant force is in a changing state, the real-time grasping state is determined to be a failure state. By analyzing the changing trend of the three-dimensional force data, the sliding direction is determined in order to plan the grasping trajectory of the dexterous hand and re-execute the grasping task. When the real-time normal resultant force is in a stable state, the real-time capture state is determined to be a state to be verified, and the corresponding stability verification action is executed.

[0012] In some implementations, performing the corresponding stability verification action includes: Control the dexterous hand to grasp the target object and lift it to a preset safe height; During the lifting process, the real-time tangential resultant force is determined by monitoring the three-dimensional force data collected by three-dimensional force sensors placed at different parts of the dexterous hand. If the real-time tangential resultant force increases to a stable value according to a preset growth law, then the real-time grasping state is determined to be a stable state. If the real-time tangential resultant force does not reach a stable value, the real-time grasping state is determined to be a slipping state, and all fingers of the dexterous hand are controlled to increase torque and the grasping task is re-executed.

[0013] In some implementations, after determining that the real-time grasping state is stable if the real-time tangential resultant force increases to a stable value according to a preset growth pattern, the method further includes: The system acquires the grasped image captured by the visual sensor and identifies the actual relative pose of the target object and the dexterous hand based on the grasped image. The actual relative pose is compared with the expected target pose of the grasping task, and the real-time grasping state is determined to be a successful grasping state when the comparison result meets the preset tolerance range.

[0014] Secondly, to achieve the above objectives, this application also proposes an electronic device, comprising: one or more processors; and a memory for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to perform the dexterous hand grasping action control method described above.

[0015] Thirdly, to achieve the above objectives, this application also proposes a computer storage medium storing executable instructions, which, when executed by a processor, cause the processor to perform the dexterous hand grasping action control method described above.

[0016] Compared with the prior art, the beneficial effects of this application include: Firstly, this application determines the diameter of the target object by analyzing its image information and compares it with multiple preset grasping diameters of the dexterous hand, thereby adaptively selecting the most suitable grasping mode (fingertip grasping mode, fingertip and second joint cooperative grasping mode, and full-hand grasping mode). This overcomes the limitations of relying on a single fixed grasping strategy in the prior art and significantly improves the grasping success rate and operational adaptability of the dexterous hand when facing objects of different sizes.

[0017] Secondly, during the grasping process, this application continuously monitors the three-dimensional force data collected by three-dimensional force sensors distributed throughout the dexterous hand, thereby sensing the grasping status in real time and triggering corresponding grasping adjustment strategies. This force-sensing-based real-time closed-loop control effectively prevents problems such as grasping failure, object slippage, or crushing, greatly enhancing the robustness of the grasping process.

[0018] Thirdly, this application deeply integrates the visual global perception (recognition and positioning) before grasping with the local fine perception of force during the grasping process, forming a collaborative system of "visual guidance decision-making and real-time force control". This enables the dexterous hand to not only "see" how to grasp, but also "feel" how well it is grasping, and make adjustments accordingly, ultimately achieving full-process intelligent control from coarse to fine. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation on the scope of this application.

[0020] Figure 1 This is a flowchart illustrating a dexterous hand grasping motion control method in one embodiment; Figure 2 This is a schematic diagram of the preset grasping diameter of a dexterous hand in one embodiment; Figure 3 This is a schematic diagram of the overall structure of a dexterous hand in one embodiment; Figure 4 This is a schematic diagram of the fingertip grasping mode in one embodiment; Figure 5 This is a schematic diagram of a grasping mode using a fingertip and a second joint in one embodiment; Figure 6 This is a schematic diagram of a grasping mode in one embodiment; Figure 7 This is a schematic diagram of the force applied to a three-dimensional force sensor arranged on a finger in one embodiment; Figure 8 This is a partial flowchart of a dexterous hand grasping action control method in one embodiment. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0022] All terms used in this application (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein should be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0023] For example, the terms "first," "second," etc., used in this application may be used herein to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish one element from another. For instance, without departing from the scope of this application, a first gripping diameter may be referred to as a second gripping diameter, and similarly, a second gripping diameter may be referred to as a first gripping diameter. Both the first gripping diameter and the second gripping diameter are gripping diameters, but they are not the same gripping diameter.

[0024] For example, the terms "comprising" or "including" used in this application indicate the presence of features, steps, operations and / or components, but do not exclude the presence or addition of one or more other features, steps, operations or components.

[0025] As mentioned earlier, current robotic dexterous hand grasping control strategies in these scenarios, especially in surgical robots, still primarily rely on pre-programmed or vision-based single action patterns. For example, they may determine the grasping point through image recognition and then execute a grasping action along a fixed trajectory. Although some systems have introduced haptic feedback, it is usually only used for simple force control or anti-slip detection, failing to effectively fuse multi-source sensor information. This results in low adaptability and success rate when facing objects of different sizes. Therefore, there is an urgent need to propose a dexterous hand grasping action control method that can adapt to task targets of different sizes. To this end, this application proposes a dexterous hand grasping action control method, electronic device, and storage medium that can improve the grasping success rate and operational adaptability of the dexterous hand when facing objects of varying sizes.

[0026] like Figure 1 As shown in the figure, this application provides a method for controlling a dexterous hand grasping action, the method including the following steps: Step S10: Receive the grabbing task and obtain the image information of the target object in the grabbing task.

[0027] In this embodiment, the grasping task refers to an instruction issued by the upper-level system or the user, requiring the robot's dexterous hand to grasp a specific target object. For example, "grab the water glass on the table." Image information refers to images, point cloud data, or combinations thereof containing the target object, obtained by capturing images of the task scene through visual sensors (such as 2D cameras, RGB-D cameras, binocular cameras, etc.) installed on the robot's head or wrist.

[0028] Step S20: By analyzing the image information, determine the target diameter corresponding to the target object.

[0029] In this embodiment, the target diameter is a parameter used to quantify a key feature of the target object's size. It does not refer to a strict geometric diameter, but rather a measure of the effective size of the target object in the grasping direction. For example, for non-cylindrical objects, the target diameter could be the diameter of the circumscribed sphere, the width of the minimum bounding box, or the maximum cross-sectional size perpendicular to the grasping direction of the dexterous hand.

[0030] In some implementations, the target object can be separated from the image information by analyzing the image information and the target diameter of the target object in a preset grasping direction can be calculated. Specifically, the target diameter can be determined by fitting the smallest circumscribed cuboid of the target object and determining the longer side of the vertical face perpendicular to the direction of the dexterous hand as the target diameter.

[0031] In some implementations, the image information is point cloud data, and the target diameter can also be selected by calculating the centroid of the point cloud data and selecting the average or maximum distance from the centroid to the surface of the point cloud data.

[0032] Step S30: Compare the target diameter with multiple preset grasping diameters of the dexterous hand, and determine the target grasping mode based on the comparison results.

[0033] In this embodiment, the preset grasping diameter refers to a series of diameter thresholds pre-set through experiments or calculations based on the mechanical structure and kinematic performance of the dexterous hand, corresponding to the optimal size range for effective operation in different grasping modes. For example... Figure 2 As shown, the multiple preset gripping diameters in this embodiment include at least a first gripping diameter A, a second gripping diameter B, and a third gripping diameter C. A gripping pattern refers to a gripping strategy designed to adapt to objects of different sizes and shapes, featuring specific finger joint movement sequences and contact point planning. For example... Figure 3 The dexterous hand shown in this embodiment includes target grasping modes such as... Figure 4 The fingertip gripping mode shown, such as Figure 5 The fingertip and second joint coordinated grasping pattern shown and such Figure 6 The fully manual grasping mode is shown.

[0034] Specifically, when the target diameter is less than or equal to the first grasping diameter of the dexterous hand, the target grasping mode is determined to be a fingertip grasping mode; when the target diameter is greater than the first grasping diameter of the dexterous hand and less than or equal to the second grasping diameter, the target grasping mode is determined to be a fingertip and second joint coordinated grasping mode; when the target diameter is greater than the second grasping diameter of the dexterous hand and less than or equal to the third grasping diameter, the target grasping mode is determined to be a full-hand grasping mode. When the target diameter is greater than the third grasping diameter of the dexterous hand, the grasping task is replanned, and the grasping direction is changed.

[0035] It should be noted that there can be more than three preset gripping diameters, and more thresholds can be introduced to define more refined gripping modes (such as side pinch, hook grip, etc.).

[0036] Step S40: Control the dexterous hand to perform the grasping task in target grasping mode. During the grasping process, the real-time grasping state is determined by monitoring the three-dimensional force data collected by three-dimensional force sensors arranged on different parts of the dexterous hand, and the corresponding grasping adjustment strategy is executed according to the real-time grasping state.

[0037] Specifically, when the target diameter is less than or equal to the first grasping diameter of the dexterous hand, the dexterous hand is controlled to perform the grasping task in a fingertip grasping mode; when the target diameter is greater than the first grasping diameter of the dexterous hand and less than or equal to the second grasping diameter, the dexterous hand is controlled to perform the grasping task in a fingertip and second joint coordinated grasping mode; when the target diameter is greater than the second grasping diameter of the dexterous hand and less than or equal to the third grasping diameter, the dexterous hand is controlled to perform the grasping task in a full-hand grasping mode.

[0038] In this embodiment, the three-dimensional force sensor refers to an array of sensing units arranged on the surface of the palm and fingers of a dexterous hand, capable of measuring pressure distribution at different locations. For example... Figure 3 As shown, the array of dots on the surface of the palm and fingers are the sensing units. Each sensing unit can decompose the multidimensional force it receives into a normal force (perpendicular to the contact surface) and a tangential force (parallel to the contact surface), thus outputting a three-dimensional force vector. The three-dimensional force sensor can be based on principles such as piezoresistive, capacitive, or optical principles; this embodiment does not limit this. Three-dimensional force data refers to a data matrix composed of the three-dimensional force vectors measured by each sensing unit in the three-dimensional force sensor at a given moment. For example... Figure 7 The gray dots shown represent active (i.e., those that have come into contact with the target object) sensing units; the darker the color, the greater the force applied.

[0039] Real-time grasping status is a dynamic evaluation of the interaction between the dexterous hand and the target object during the grasping process, which can include failure status, pending verification status, stable status, slippage status, and successful grasping status. The grasping adjustment strategy is a set of predefined, adaptive control instructions corresponding to the real-time grasping status.

[0040] like Figure 8 As shown, in step S40, the real-time grasping state is determined by monitoring the three-dimensional force data collected by three-dimensional force sensors arranged at different parts of the dexterous hand, and a corresponding grasping adjustment strategy is executed according to the real-time grasping state, including: Step S41: Determine the real-time normal resultant force by monitoring the three-dimensional force data collected by three-dimensional force sensors arranged at different parts of the dexterous hand.

[0041] In this embodiment, the real-time normal resultant force refers to the scalar sum of the normal forces (i.e., gripping pressures) perpendicular to their respective contact surfaces measured by all activated (i.e., those that have contacted the target object) three-dimensional force sensors at a certain moment, used to quantify the total gripping force applied to the target object by the dexterous hand.

[0042] Step S42: When the real-time normal resultant force is in a changing state, the real-time grasping state is determined to be a failure state. By analyzing the changing trend of the three-dimensional force data, the sliding direction is determined in order to plan the grasping trajectory of the dexterous hand and re-execute the grasping task.

[0043] In this embodiment, a changing state refers to a situation where the real-time normal resultant force is in an unstable fluctuation or a continuously decreasing trend, which means that there is relative motion (i.e., sliding) between the target object and the dexterous hand. A failure state refers to an unsuccessful grasp, corresponding to the specific grasping anomaly of "sliding" represented by the changing state of the real-time normal resultant force.

[0044] The sliding direction refers to the overall motion trend of the target object relative to the dexterous hand, inferred by analyzing the distribution trend of the sensing units in contact with the target object in the three-dimensional force sensor over time. For example, at the first moment, the pressure center of the sensing unit in contact with the target object is at the first position coordinate, and at the second moment, the pressure center of the sensing unit in contact with the target object is at the second position coordinate. Therefore, from the first moment to the second moment, the target object slides from the first position coordinate to the second position coordinate relative to the dexterous hand, and the sliding direction is from the first position coordinate to the second position coordinate.

[0045] Furthermore, a corresponding compensation trajectory can be generated based on the sliding direction. For example, the dexterous hand or robotic arm that has grasped the target object can be controlled to make a small compliant movement along the sliding direction to complete the pose compensation between the dexterous hand and the target object. After completing the pose compensation, the grasping action is executed again to stabilize the target object.

[0046] Step S43: When the real-time normal resultant force is in a stable state, determine the real-time grasping state as a state to be verified, and execute the corresponding stability verification action.

[0047] In this embodiment, a stable state refers to a state where the real-time normal resultant force remains constant within a small threshold range after reaching a preset value, without any detected significant decrease or abnormal fluctuation. The state to be verified is an intermediate state, indicating that the initial force perception test has passed, but whether the grasping is truly stable still requires further verification. The stability verification action refers to a series of exploratory actions and judgment logic actively executed to test whether the grasping is sufficiently stable.

[0048] In some implementations, stability can be verified by performing a lifting test. Specifically, a dexterous hand can be controlled to grasp a target object and lift it to a preset safe height. The preset safe height refers to a pre-set, relatively small lifting distance (e.g., 2-5 cm). This height needs to be small enough that even if the target object slips, it will not cause serious damage or danger; at the same time, it needs to be large enough to effectively simulate the inertial forces at the beginning of movement, sufficient to overcome static friction and trigger potential slippage. Furthermore, the preset safe height can be dynamically set according to the estimated weight and surface material of the target object, and is inversely proportional to both the estimated weight and smoothness of the target object.

[0049] During the lifting process, the real-time tangential resultant force is determined by monitoring the three-dimensional force data collected by three-dimensional force sensors deployed at different parts of the dexterous hand. The real-time tangential resultant force refers to the vector sum of the tangential forces (frictional forces) measured by all activated sensing units during the lifting process, which are parallel to the contact surface and opposite in direction to gravity. In vertical lifting scenarios, it can specifically refer to the sum of the vertically downward tangential forces measured by all sensing units, reflecting the total static friction force provided by the dexterous hand to resist the gravity of the target object.

[0050] If the real-time tangential resultant force increases to a stable value according to a preset growth pattern, then the real-time grasping state is determined to be a stable state. The preset growth pattern refers to the ideal change pattern that the real-time tangential resultant force should exhibit during the lifting process. This can be characterized by starting from zero or a low value and increasing smoothly and rapidly with the application of lifting acceleration and the full effect of gravity. The stable value refers to the state in which the real-time tangential resultant force maintains a small fluctuation around a value equal to or slightly greater than the object's weight when the lifting enters a uniform speed phase or a brief pause.

[0051] In some implementations, after determining that the real-time grasping state is a stable state, the grasped image collected by the visual sensor can be acquired, and based on the grasped image, the actual relative pose of the target object and the dexterous hand can be identified; the actual relative pose is compared with the expected target pose of the grasping task, and when the comparison result meets the preset tolerance range, the real-time grasping state is determined to be a successful grasping state.

[0052] The post-grasp image refers to the image data containing the grasped object and the dexterous hand's end effector, which is collected again by the vision sensor after force verification is passed. It reflects the final state after the grasping action is completed. The actual relative pose refers to the relative position and posture relationship between the target object and the dexterous hand in three-dimensional space, calculated from the post-grasp image using image recognition algorithms. Examples include the target object being completely enveloped by the palm, the target object's long axis being parallel to the middle finger, or specific feature points of the target object being located at the center of the line connecting the thumb and index finger. The expected target pose refers to the ideal relative pose that is predefined according to the requirements of the grasping task and is expected to be achieved between the target object and the dexterous hand. The preset tolerance range refers to a predefined maximum acceptable deviation in position (e.g., ±1 cm) and direction (e.g., ±5 degrees) between the actual relative pose and the expected target pose.

[0053] This implementation proposes visual closed-loop verification based on force-based closed-loop verification, which can prevent subsequent operations (such as assembly and placement) from failing due to incorrect posture despite a firm grip. For example, in the task of grasping a scalpel, the expected target posture is that the blade is parallel to the index finger (0-degree angle), and the end of the handle is in the palm. The system calculates the actual relative posture (a 5-degree angle between the blade and the index finger, and the end of the handle in the palm) and compares it with the expected posture. If the 5-degree angle error is determined to be within the preset ±5-degree tolerance range, the grasp is considered successful, and subsequent suturing operations can proceed.

[0054] If the real-time tangential resultant force fails to reach a stable value, the real-time grasping state is determined to be a slipping state. The dexterous hand is then controlled to increase torque across all fingers and the grasping task is re-executed. The slipping state refers to the inability of the real-time tangential resultant force to increase and maintain a stable value during the lifting process. This may manifest as slow force growth, premature descent, or continuous decay. This indicates that the target object is slipping relative to the dexterous hand. Increasing torque involves increasing the output of all finger drive motors to enhance normal pressure, thereby increasing the maximum static friction. After increasing the torque, the dexterous hand can be controlled to re-execute the grasping task, for example, returning to a preset grasping position to repeat the grasping and verification cycle.

[0055] Through steps S41-S43, this application constructs a two-level grasping state discrimination and response mechanism. Specifically, firstly, by monitoring the real-time normal resultant force, the real-time grasping state is quickly divided into two categories: failure state and unverified state. For the failure state, this application analyzes real-time three-dimensional force data to determine the sliding direction and actively performs trajectory compensation. For the unverified state, this application effectively identifies "false grasping" that appears stable under static conditions but fails upon disturbance through stability verification actions. For example, when grasping smooth or oily objects, slow slippage may occur even if the real-time normal resultant force is sufficient. Once the real-time tangential resultant force is detected to be unstable (i.e., determined to be a slippage state), the system can automatically strengthen the grasping force and attempt again, thereby significantly improving the success rate and robustness of the entire system when facing uncertain object properties (such as material and weight).

[0056] In the dexterous hand grasping action control method proposed in this embodiment, in the first aspect, this application determines the diameter of the target object by analyzing the image information of the target object and compares it with multiple preset grasping diameters of the dexterous hand, thereby adaptively selecting the most suitable grasping mode (fingertip grasping mode, fingertip and second joint cooperative grasping mode, and full hand grasping mode). This overcomes the limitation of relying on a single fixed grasping strategy in the prior art and significantly improves the grasping success rate and operational adaptability of the dexterous hand when facing objects of different sizes.

[0057] Secondly, during the grasping process, this application continuously monitors the three-dimensional force data collected by three-dimensional force sensors distributed throughout the dexterous hand, thereby sensing the grasping status in real time and triggering corresponding grasping adjustment strategies. This force-sensing-based real-time closed-loop control effectively prevents problems such as grasping failure, object slippage, or crushing, greatly enhancing the robustness of the grasping process.

[0058] Thirdly, this application deeply integrates the visual global perception (recognition and positioning) before grasping with the local fine perception of force during the grasping process, forming a collaborative system of "visual guidance decision-making and real-time force control". This enables the dexterous hand to not only "see" how to grasp, but also "feel" how well it is grasping, and make adjustments accordingly, ultimately achieving full-process intelligent control from coarse to fine.

[0059] In one embodiment, the target grasping mode is a fingertip grasping mode, and step S40, controlling the dexterous hand to perform the grasping task in the target grasping mode, includes: Step A10: Control the dexterous hand to move to the preset grasping position with the thumb and forefinger spread apart.

[0060] In this embodiment, the open posture of the thumb and index finger refers to... Figure 4As shown, the thumb and forefinger are positioned far apart, forming a "C" or V-shaped pre-grasping space to ensure no collision occurs when approaching the target object and to allow for gripping from opposite sides of the object. The preset gripping position is a spatial point calculated by the system based on the target object's location before performing the gripping operation. This position ensures that when the dexterous hand moves to this position, the outstretched thumb and forefinger can accurately close from both sides of the target object.

[0061] Step A20: Control the thumb and index finger to perform a pinching action, and stop the pinching action of the contacting fingers when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object.

[0062] In this embodiment, the pinching action refers to the movement of the thumb and index finger, as a pair of antagonistic fingers, approaching each other to grasp the target object. While the thumb and index finger are performing the pinching action, the three-dimensional force sensor data on the fingertips of the thumb and index finger can be read in real time. Once the force value of the three-dimensional force sensor on either fingertip (e.g., the thumb) in the normal direction (perpendicular to the contact surface) exceeds a small contact threshold (e.g., 0.1N), the movement of that finger (thumb) immediately stops. The other finger (index finger) continues to move until its fingertip's three-dimensional force sensor also detects contact force, and then stops. This achieves compliant control using force feedback, preventing the object from being pushed over or crushed due to inaccurate position control.

[0063] Step A30: Control other fingers to perform a closing action towards the target object, and stop performing the closing action of the contacting fingers when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object.

[0064] In this embodiment, "other fingers" refers to fingers other than the thumb and index finger, such as... Figure 3 In the diagram of the five fingers in a dexterous hand, the middle, ring, and little fingers are used. The closing motion refers to the other fingers bending towards the palm, approaching and contacting the target object from the side or below to provide additional support points and prevent the object from rotating or slipping. By using the closing motion of the other fingers, more contact points can be formed, transforming a two-finger pinch into a multi-finger envelopment, effectively preventing the object from rotating or slipping during subsequent operations (such as moving or placing), greatly enhancing the system's ability to resist disturbances.

[0065] Step A40: Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

[0066] In this embodiment, increasing torque refers to the system outputting a larger current or command to the joint motors of each finger through the dexterous hand controller, thereby increasing the gripping force exerted by each finger on the object. The normal resultant force refers to the scalar sum of the forces (i.e., gripping forces) perpendicular to the contact surface measured by all three-dimensional force sensors on the dexterous hand, representing the total clamping force of the entire hand on the target object. The preset value is an optimal gripping force threshold pre-set through experiments or calculations to ensure a firm grip without damaging the object. It can be dynamically queried and adjusted based on the estimated material (through image recognition) and weight of the target object.

[0067] The dexterous hand grasping action control method proposed in this embodiment achieves a highly intelligent, adaptive and robust grasping process by starting with precise positioning of force triggering by the dominant fingers (thumb and index finger), then coordinating and stabilizing force triggering by the auxiliary fingers (other fingers), and finally realizing reliable force closure through force closed-loop control.

[0068] In one embodiment, the target grasping mode is a fingertip and second joint coordinated grasping mode, and step S40, controlling the dexterous hand to perform the grasping task in the target grasping mode, includes: Step B10: Control the dexterous hand to move to the preset grasping position with the second joints of the thumb and index finger spread to the first preset angle and the fingertips spread to the preset limit angle.

[0069] In this embodiment, the first preset angle refers to a preset, non-zero joint angle used to keep the second joints of the thumb and index finger open. The preset limit angle refers to the maximum angle at which the fingertips can open, thereby maximizing space before the second joints contact the target object and preventing the fingertips from touching the object prematurely. The preset grasping position refers to a spatial point calculated by the system based on the position of the target object before performing the grasping operation. This position ensures that when the dexterous hand moves to this position, the open thumb and index finger can accurately close from both sides of the target object.

[0070] Step B20: Sequentially control the second joints of each finger of the dexterous hand to perform tightening actions, and stop performing the tightening action of the contacting finger when the three-dimensional force sensor of the corresponding second joint detects the contact force with the target object.

[0071] In this embodiment, the tightening action refers to the bending movement of the second joint, which is used to bring the fingertips inward.

[0072] In one implementation, the second joint motors of all fingers can be driven to perform a tightening action. Once the three-dimensional force sensor corresponding to the second joint of a finger detects a contact force (e.g., >0.1N), the tightening of that finger's second joint is immediately stopped. The other uncontacted fingers continue to move until the fingertips of the second joints of all fingers are in contact with the surface of the object.

[0073] In another embodiment, the second joint motors of the thumb and index finger can be driven first to perform a tightening action. Once the three-dimensional force sensor corresponding to the second joint of any finger (e.g., the thumb) detects a contact force (e.g., >0.1N), the movement of that finger (thumb) is immediately stopped. The other finger (index finger) continues to move until the three-dimensional force sensor corresponding to its second joint also detects a contact force, and then stops. Next, the second joint motors of the other fingers are driven to perform a tightening action. Once the three-dimensional force sensor corresponding to the second joint of any finger detects a contact force (e.g., >0.1N), the tightening of that finger's second joint is immediately stopped. The other fingers that have not yet made contact continue to move until the fingertips of the second joints of all fingers are in contact with the surface of the object.

[0074] This implementation prioritizes the thumb and index finger, a key pair of antagonistic fingers, to make initial contact with the target object. This rapidly establishes stable and reliable mechanical fulcrums on both sides of the object, effectively preventing the object from rolling or shifting due to uneven force distribution during the subsequent envelopment process of the remaining fingers. Then, other fingers are controlled to contact the target object, allowing the system to transition more smoothly and controllably to a fully enveloping state. This strategy reduces the high requirements for the synchronization accuracy of all finger joint trajectories and enhances the system's tolerance to minor uncertainties in the actual pose of the object, thereby achieving a higher success rate and more stable grasping initialization in complex scenarios.

[0075] Step B30: Sequentially control the fingertips of each finger of the dexterous hand to perform a clamping action, and stop the clamping action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object.

[0076] In this embodiment, the clamping action refers to the bending motion of the fingertips, causing them to eventually contact and press against the target object. Similarly, the fingertip joint motors of all fingers can be driven to perform the clamping action. Once the three-dimensional force sensor corresponding to the fingertip of a certain finger detects a contact force (e.g., >0.1N), the clamping action of that fingertip is immediately stopped. The other fingers that have not yet made contact continue to move until the fingertips of all fingers are in contact with the object surface. Alternatively, the fingertip motors of the thumb and index finger can be driven first to perform the clamping action. Once the three-dimensional force sensor corresponding to the fingertip of any finger (e.g., the thumb) detects a contact force (e.g., >0.1N), the movement of that finger (thumb) is immediately stopped. The other finger (index finger) continues to move until the three-dimensional force sensor corresponding to its fingertip also detects a contact force, and then stops. Then the fingertip motors of the other fingers are driven to perform the clamping action. Once the three-dimensional force sensor corresponding to the fingertip of a certain finger detects a contact force (e.g., >0.1N), the clamping action of that fingertip is immediately stopped. The other fingers that are not in contact continue to move until the tips of all fingers are in contact with the surface of the object.

[0077] Step B40: Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

[0078] In this embodiment, increasing torque refers to the system outputting a larger current or command to the joint motors of each finger through the dexterous hand controller, thereby increasing the gripping force exerted by each finger on the object. The normal resultant force refers to the scalar sum of the forces (i.e., gripping forces) perpendicular to the contact surface measured by all three-dimensional force sensors on the dexterous hand, representing the total clamping force of the entire hand on the target object. The preset value is an optimal gripping force threshold pre-set through experiments or calculations to ensure a firm grip without damaging the object. It can be dynamically queried and adjusted based on the estimated material (through image recognition) and weight of the target object.

[0079] In the dexterous hand grasping motion control method proposed in this embodiment, the grasping process is decomposed into two stages, "stable support" and "fine constraint", by using phased and part-sequential contact control (first the second joint fingertip, then the fingertip). This fully utilizes the structural characteristics of different parts of the dexterous hand and achieves a grasping process from coarse to fine through force feedback. It perfectly solves the problem of stable and adaptive grasping of medium-sized objects, especially spherical or easily rolling objects, and significantly improves the success rate and robustness of grasping.

[0080] In one embodiment, the target grasping mode is a full-hand grasping mode, and step S40, controlling the dexterous hand to perform the grasping task in the target grasping mode, includes: Step C10: Control the dexterous hand to move toward the target object with all fingers' third joints open to a second preset angle, and the second knuckles and fingertips open to a preset limit angle, until the movement stops when the three-dimensional force sensor arranged in the palm detects the contact force.

[0081] In this embodiment, the second preset angle refers to an angle set to allow the palm to cover a large object, enabling all third joints to fully open. The preset limit angle refers to the maximum angle at which the second joints and fingertips can extend, in order to prevent the second joints and fingertips from touching the target object prematurely.

[0082] By stopping movement when a three-dimensional force sensor located on the palm detects a contact force, the palm can be used as the initial anchor point for the grasping process, preventing the target object from being pushed away when the fingers subsequently close.

[0083] Step C20: Sequentially control the third joint of each finger of the dexterous hand to perform a tightening action, and stop the tightening action of the contacting finger when the three-dimensional force sensor of the corresponding third joint detects the contact force with the target object.

[0084] In one implementation, the motors of the third joints of all fingers can be driven to perform a tightening action. Once the three-dimensional force sensor corresponding to the third joint of a finger detects a contact force (e.g., >0.1N), the tightening of the third joint of that finger is immediately stopped. The other uncontacted fingers continue to move until the fingertips of the third joints of all fingers are in contact with the surface of the object.

[0085] In another embodiment, the third joint motors of the thumb and index finger can be driven first to perform a tightening action. Once the three-dimensional force sensor corresponding to the third joint of any finger (e.g., the thumb) detects a contact force (e.g., >0.1N), the movement of that finger (thumb) is immediately stopped. The other finger (index finger) continues to move until the three-dimensional force sensor corresponding to its third joint also detects a contact force, and then stops. Next, the third joint motors of the other fingers are driven to perform a tightening action. Once the three-dimensional force sensor corresponding to the third joint of any finger detects a contact force (e.g., >0.1N), the tightening of that finger's third joint is immediately stopped. The other fingers that have not yet made contact continue to move until the fingertips of the third joints of all fingers are in contact with the surface of the object.

[0086] By tightening the third joint, the initial envelopment and constraint of the main body of the target object can be quickly achieved.

[0087] In step C30, the second joints of each finger of the dexterous hand are sequentially controlled to perform tightening actions, and the tightening action of the contacting finger is stopped when the three-dimensional force sensor of the corresponding second joint detects the contact force with the target object.

[0088] In one implementation, the second joint motors of all fingers can be driven to perform a tightening action. Once the three-dimensional force sensor corresponding to the second joint of a finger detects a contact force (e.g., >0.1N), the tightening of that finger's second joint is immediately stopped. The other uncontacted fingers continue to move until the fingertips of the second joints of all fingers are in contact with the surface of the object.

[0089] In another embodiment, the second joint motors of the thumb and index finger can be driven first to perform a tightening action. Once the three-dimensional force sensor corresponding to the second joint of any finger (e.g., the thumb) detects a contact force (e.g., >0.1N), the movement of that finger (thumb) is immediately stopped. The other finger (index finger) continues to move until the three-dimensional force sensor corresponding to its second joint also detects a contact force, and then stops. Next, the second joint motors of the other fingers are driven to perform a tightening action. Once the three-dimensional force sensor corresponding to the second joint of any finger detects a contact force (e.g., >0.1N), the tightening of that finger's second joint is immediately stopped. The other fingers that have not yet made contact continue to move until the fingertips of the second joints of all fingers are in contact with the surface of the object.

[0090] By tightening the second joint, a second layer of contact constraint is added on the basis of the initial envelope, which makes the shape of the dexterous hand better adapt to the geometry of the target object and significantly increases the contact area and friction points.

[0091] Step C40: Sequentially control the fingertips of each finger of the dexterous hand to perform a clamping action, and stop the clamping action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object.

[0092] In this embodiment, the clamping action refers to the bending motion of the fingertips, causing them to eventually contact and press against the target object. Similarly, the fingertip joint motors of all fingers can be driven to perform the clamping action. Once the three-dimensional force sensor corresponding to the fingertip of a certain finger detects a contact force (e.g., >0.1N), the clamping action of that fingertip is immediately stopped. The other fingers that have not yet made contact continue to move until the fingertips of all fingers are in contact with the object surface. Alternatively, the fingertip motors of the thumb and index finger can be driven first to perform the clamping action. Once the three-dimensional force sensor corresponding to the fingertip of any finger (e.g., the thumb) detects a contact force (e.g., >0.1N), the movement of that finger (thumb) is immediately stopped. The other finger (index finger) continues to move until the three-dimensional force sensor corresponding to its fingertip also detects a contact force, and then stops. Then the fingertip motors of the other fingers are driven to perform the clamping action. Once the three-dimensional force sensor corresponding to the fingertip of a certain finger detects a contact force (e.g., >0.1N), the clamping action of that fingertip is immediately stopped. The other fingers that are not in contact continue to move until the tips of all fingers are in contact with the surface of the object.

[0093] The final closure of the grasping form is achieved through the clamping motion of the fingertips, forming a complete force flow path from the palm to the fingertips, which can prevent the target object from slipping off the top when subjected to complex external forces.

[0094] Step C50: Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

[0095] In this embodiment, increasing torque refers to the system outputting a larger current or command to the joint motors of each finger through the dexterous hand controller, thereby increasing the gripping force exerted by each finger on the object. The normal resultant force refers to the scalar sum of the forces (i.e., gripping forces) perpendicular to the contact surface measured by all three-dimensional force sensors on the dexterous hand, representing the total clamping force of the entire hand on the target object. The preset value is an optimal gripping force threshold pre-set through experiments or calculations to ensure a firm grip without damaging the object. It can be dynamically queried and adjusted based on the estimated material (through image recognition) and weight of the target object.

[0096] In the dexterous hand grasping motion control method proposed in this embodiment, firstly, by strictly layering sequential contact control from proximal to distal (palm → third joint → second joint → fingertip), the grasping process is decomposed into multiple progressive, sensor-driven stages, achieving a progressive, adaptive envelope for large objects. Secondly, through the stable base established by the palm and the progressively increasing constraints of the fingers, this application forms a grasping state with the largest contact area, optimal force distribution, and extremely high stability, greatly improving the ability and reliability of the dexterous hand to grasp large and heavy objects.

[0097] In one embodiment, a computer storage medium is provided that stores executable instructions that, when executed by a processor, cause the processor to perform the steps in the above method embodiments.

[0098] In one embodiment, an electronic device is also provided, including one or more processors; and a memory storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors perform the steps in the above method embodiments.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

[0100] Furthermore, those skilled in the art will understand that although some embodiments herein include certain features included in other embodiments but not others, combinations of features from different embodiments are intended to be within the scope of this application and form different embodiments. For example, any of the embodiments or implementations claimed above can be used in any combination. The information disclosed in this background section is intended only to enhance the understanding of the general background of this application and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

Claims

1. A method for controlling the grasping action of a dexterous hand, characterized in that, The method includes: Receive a grabbing task and acquire image information of the target object in the grabbing task; By analyzing the image information, the target diameter corresponding to the target object is determined; The target diameter is compared with multiple preset grasping diameters of a dexterous hand, and the target grasping mode is determined based on the comparison results. The target grasping mode includes a fingertip grasping mode, a fingertip and second joint coordinated grasping mode, and a full-hand grasping mode. The dexterous hand is controlled to perform the grasping task in a target grasping mode. During the grasping process, the real-time grasping state is determined by monitoring the three-dimensional force data collected by three-dimensional force sensors arranged on different parts of the dexterous hand, and the corresponding grasping adjustment strategy is executed according to the real-time grasping state.

2. The dexterous hand grasping action control method according to claim 1, characterized in that, The plurality of preset grasping diameters includes at least a first grasping diameter, a second grasping diameter, and a third grasping diameter; the step of comparing the target diameter with the plurality of preset grasping diameters of the dexterous hand and determining the target grasping pattern based on the comparison result includes: When the target diameter is less than or equal to the first grasping diameter of the dexterous hand, the dexterous hand is controlled to perform the grasping task in a fingertip grasping mode. When the target diameter is greater than the first grasping diameter of the dexterous hand and less than or equal to the second grasping diameter, the dexterous hand is controlled to perform the grasping task in a grasping mode that coordinates the fingertips and the second joint. When the target diameter is greater than the second grasping diameter of the dexterous hand and less than or equal to the third grasping diameter, the dexterous hand is controlled to perform the grasping task in a full-hand grasping mode.

3. The dexterous hand grasping motion control method according to claim 2, characterized in that, The controlled dexterous hand performs the grasping task in a fingertip grasping mode, including: Control the dexterous hand to move to the preset grasping position with the thumb and forefinger spread out; The thumb and index finger are controlled to perform a pinching action, and the pinching action of the contacting fingers is stopped when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object. Control other fingers to perform a retraction action towards the target object, and stop performing the retraction action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object; Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

4. The dexterous hand grasping action control method according to claim 2, characterized in that, The controlled dexterous hand performs the grasping task using a fingertip and second joint coordinated grasping mode, including: Control the dexterous hand to move to the preset grasping position with the second joints of the thumb and index finger spread to the first preset angle and the fingertips spread to the preset limit angle; The second joints of each finger of the dexterous hand are controlled in sequence to perform tightening actions, and the tightening action of the contacting finger is stopped when the three-dimensional force sensor of the corresponding second joint detects the contact force with the target object. The dexterous hand sequentially controls the fingertips of each finger to perform a clamping action, and stops performing the clamping action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object; Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

5. The dexterous hand grasping action control method according to claim 2, characterized in that, The controlled dexterous hand performs the grasping task in a full-hand grasping mode, including: Control the dexterous hand to move toward the target object with all fingers' third joints spread to a second preset angle, and the second knuckles and fingertips spread to a preset limit angle, until the movement stops when the three-dimensional force sensor arranged in the palm detects the contact force; The third joint of each finger of the dexterous hand is controlled to perform a tightening action in sequence, and the tightening action of the contacting finger is stopped when the three-dimensional force sensor of the corresponding third joint detects the contact force with the target object. The second joints of each finger of the dexterous hand are controlled in sequence to perform tightening actions, and the tightening action of the contacting finger is stopped when the three-dimensional force sensor of the corresponding second joint detects the contact force with the target object. The dexterous hand sequentially controls the fingertips of each finger to perform a clamping action, and stops performing the clamping action of the contacting finger when the three-dimensional force sensor of the corresponding fingertip detects the contact force with the target object; Control all fingers of the dexterous hand to increase torque to grasp the target object until the resultant normal force detected by all three-dimensional force sensors reaches a preset value.

6. The dexterous hand grasping action control method according to claim 1, characterized in that, The process of determining the real-time grasping state by monitoring three-dimensional force data collected by three-dimensional force sensors deployed at different parts of the dexterous hand, and executing corresponding grasping adjustment strategies based on the real-time grasping state, includes: By monitoring the three-dimensional force data collected by three-dimensional force sensors placed at different parts of the dexterous hand, the real-time normal resultant force is determined; When the real-time normal resultant force is in a changing state, the real-time grasping state is determined to be a failure state. By analyzing the changing trend of the three-dimensional force data, the sliding direction is determined in order to plan the grasping trajectory of the dexterous hand and re-execute the grasping task. When the real-time normal resultant force is in a stable state, the real-time capture state is determined to be a state to be verified, and the corresponding stability verification action is executed.

7. The dexterous hand grasping motion control method according to claim 6, characterized in that, The execution of the corresponding stability verification action includes: Control the dexterous hand to grasp the target object and lift it to a preset safe height; During the lifting process, the real-time tangential resultant force is determined by monitoring the three-dimensional force data collected by three-dimensional force sensors placed at different parts of the dexterous hand. If the real-time tangential resultant force increases to a stable value according to a preset growth law, then the real-time grasping state is determined to be a stable state. If the real-time tangential resultant force does not reach a stable value, the real-time grasping state is determined to be a slipping state, and all fingers of the dexterous hand are controlled to increase torque and the grasping task is re-executed.

8. The dexterous hand grasping action control method according to claim 7, characterized in that, After determining that the real-time grasping state is stable if the real-time tangential resultant force increases to a stable value according to a preset growth law, the method further includes: The system acquires the grasped image captured by the visual sensor and identifies the actual relative pose of the target object and the dexterous hand based on the grasped image. The actual relative pose is compared with the expected target pose of the grasping task, and the real-time grasping state is determined to be a successful grasping state when the comparison result meets the preset tolerance range.

9. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors perform the dexterous hand grasping motion control method as described in any one of claims 1 to 8.

10. A computer storage medium, characterized in that, The storage medium stores executable instructions, which, when executed by a processor, cause the processor to perform the dexterous hand grasping motion control method as described in any one of claims 1 to 8.

Images